Questions about Conversations with a System

My thoughts and questions prompted by an article today In the NYT:  As alert systems and devices start having extended conversations with us, can we expect them to be affirming?  Critical?  Congratulatory?   I am working on a system that will provide groups of users with critique of their work.  To what degree should that system do these things, even in a subtle way?  You can think of a congratulatory response as a form of gamification, especially when there is competition involved.  But how about affirmation?  Or criticism?  How are memory or trending included in such an assessment?  And, are we already at this point, or should we wait until the sensors can do a better job of inferring quality?   Thoughts welcome.